Learn
Bring your own AI keys: when it makes sense
Every Somebody plan includes avatar minutes with Included AI. You can also run agents on your own OpenAI, Google Gemini or Azure OpenAI account. Here is how to choose between the two.
Every Somebody agent needs a language model to understand what people say and decide what to say back. There are two ways to provide one. You can use the included AI that comes with your plan, or you can bring your own keys from an AI provider you already have an account with.
Neither is the better choice in general. They suit different teams and different stages. This guide sets out what changes when you switch, and how to decide.
Option 1: included AI
Included AI is the simplest path. You create an agent, choose an avatar and a voice, and it works. There is no provider account to set up and no keys to manage. Usage is measured in minutes of conversation and drawn from your plan's monthly allowance.
- Free: 30 avatar minutes a month with Included AI.
- Pro: 150 avatar minutes a month with Included AI, then $0.12 per minute.
- Team: 2,000 avatar minutes a month with Included AI.
Included AI is a good fit when you are getting started, testing an idea, running a modest volume of conversations, or simply do not want another vendor relationship.
Option 2: your own keys
With your own keys, you connect an account you hold with OpenAI, Google Gemini or Azure OpenAI. Your agents then send their AI requests through that account. The provider bills you directly for model usage, and Somebody charges a lower per-minute rate for everything else: the face, the voice pipeline, the platform.
- Free: up to 60 avatar minutes a day with your own AI keys, with conversations capped at 3 minutes.
- Pro: 1,000 avatar minutes a month with your own AI keys, then $0.03 per minute.
- Team: 20,000 avatar minutes a month with your own AI keys.
Full details are on the pricing page.
Why teams bring their own keys
Control
Your keys mean your provider account. You choose the provider and see usage in your own dashboard. If your organisation has standardised on one provider, own keys keep your agents inside that decision rather than adding a new one.
Provider agreements you already have
Many organisations have already negotiated terms with an AI provider, covering data handling, retention, region or pricing. Azure OpenAI in particular is often chosen because it sits inside an existing cloud agreement. Bringing your own keys lets your agents run under those terms.
To be clear about what this does and does not do: your keys route the model requests through your account. The rest of the agent, including the face and the conversation itself, still runs on Somebody. Own keys are a way to use your provider agreement, not a way to remove Somebody from the picture.
Cost at scale
At low volume, the difference rarely matters, and included AI is simpler. As volume grows, the arithmetic can change. Take the Pro plan's public prices as an example. Beyond the included allowance, an included-AI minute costs $0.12. Beyond the own-key allowance, an own-key minute costs $0.03, plus whatever your provider charges for the model usage in that minute.
Whether own keys come out cheaper depends on your provider's rates, the model you choose and how talkative your conversations are. We cannot give you a single answer, because the provider half of the bill is between you and them. What we can say is that the Somebody half is lower on own keys, and the allowances are much larger: 1,000 avatar minutes with your own AI keys against 150 with Included AI on Pro, and 20,000 against 2,000 on Team.
Run a week on included AI, look at your minutes, then price the same volume on your provider's rate card. That gives you a real comparison instead of a guess.
Why teams stay on included AI
- One bill. No second invoice to reconcile and no provider spend to monitor.
- No setup. No account, no billing on the provider side, no key rotation.
- Predictable. You know the per-minute price up front.
- Fewer moving parts. No provider quota or spending limit on your account that might stop agents mid-day.
Your responsibilities on your own keys
Bringing your own keys moves some responsibility to you, and it is worth being plain about it.
- Provider terms apply to you. Your use of the model is governed by your agreement with that provider, including its usage policies. Make sure your use case is allowed under them.
- Provider billing is yours. The provider charges your account for model usage. Set spending limits and alerts there.
- Quotas and availability. If your provider account hits a rate limit or a spending cap, your agents will be affected. Check your limits before a launch.
- Key hygiene. Use a key created for this purpose, so you can rotate or revoke it without disrupting anything else.
How to decide
A few questions usually settle it.
- Does your organisation require a specific AI provider or agreement? If yes, bring your own keys.
- Are you still exploring? Start on included AI. You can move to your own keys once you know what you need.
- Are you running a lot of conversations? Measure your real usage, then compare included-AI pricing with own-key pricing plus your provider's rates.
- Who will own the provider account? If nobody wants to manage keys, limits and a second bill, included AI is the honest answer.
Switching is not a one-way door
The choice is not permanent. A sensible path is to start on included AI, learn what your conversations look like, and move to your own keys once volume or policy makes the case, or go the other way when simplicity matters more.
To see how the minutes and rates compare plan by plan, visit the pricing page. If you would like to talk it through, book a demo and we will help you work out which fits.